Research

Classical, Hybrid Quantum-Classical, and Quantum Multi-Agent Deep Reinforcement Learning for Optimal Resource Allocation in AoI-Aware Energy-Efficient Platoon-Based C-V2X Systems

This work aims to tackle the complex problem of channel assignment and joint power-energy allocation within a cellular-vehicle-to-everything (C-V2X) network in an urban traffic intersection. Here, the C-V2X network is strategically deployed to facilitate the coordination of multiple vehicle platoons. This includes updating platoon states to the roadside unit (RSU) and managing the exchange of cooperative awareness messages (CAMs) among vehicles. Our objective is to minimise the average age of information (AoI), maximise the CAM delivery probability, and foster sustainable, green communication strategies through efficient management for both power and energy.  Given the intricate nature of this challenge, we adopt a multi-agent deep reinforcement learning (MADRL) approach based on Markov decision process (MDP). Next, we introduce two innovative algorithms based on multi-agent deep deterministic policy gradient (MADDPG) and twin delayed deep deterministic policy gradient (TDDD) to effectively address the optimisation problem. Finally, the simulation results demonstrate remarkable performance in terms of energy efficiency, while maintaining algorithm convergence speed and AoI level.

In addition to the above classical MADRL approach, we investigate the potential advantages of integrating quantum neural networks (QNNs) with classical MADRL approaches. Specifically, we combine a variational quantum circuit (VQC) with traditional neural networks to develop a hybrid quantum-classical neural network for the MADRL training process. Compared with classical approaches, our numerical results reveal that the proposed method achieves exceptional energy efficiency performance, while maintaining the algorithm convergence rate and AoI levels.

The next step of this work will focus on developing better quantum algorithms to reach a lower AoI level within more complicated vehicle systems while still keeping the sustainable green communication strategy. In addition, instead of using the quantum simulator, real quantum computers, such as IBM quantum computers, are considered to be used in this research.

Resolving High Dynamic Range Time-of-Flight Imaging with Multipath Interference

The saturation of analog-to-digital converter (ADC) and multipath interference (MPI) are two of the main problems in Time-of-Flight (TOF) imaging. The former leads to the clipping of the received high dynamic range (HDR) signals which results in the loss of information and hence incorrect estimation of TOF parameters: amplitude and depth. The latter leads to the superposition of multiple reflected signals at a single pixel which corrupts the parameters, hence also increases the inaccuracy in the estimated amplitude and depth. This work aims at tackling the joint problem of the HDR signals and the MPI issue. The model in this work considers the input HDR signal to be composed of L paths with different amplitudes and depths at each pixel of the TOF image. The dynamic range of the ADC is [−λ,λ] and the signal dynamic range can be twenty times larger than this. Based on the Unlimited Sensing Framework (USF), the TOF parameters are extracted from the modulo samples of the HDR signal under the conditions for recovery stated by the USF. Then, the MPI is resolved using the classical parametric methods for line spectra by taking multi-frequency TOF measurements. The numerical experiment shows the feasibility of this approach. 

Simultaneous Wireless Computing and Power Transfer in Fog Radio Access Network

This research considers a Fog Radio Access Network (F-RAN) where the low-power communication devices (energy users, EUs) can harvest energy from radiofrequency (RF) signal, and the enhanced Remote Radio Heads (eRRHs) can execute computationally demanding tasks offloaded by resource- and power-constrained mobile content users (CUs). This research tackles the problem of computation offloading and energy harvesting in this F-RAN by jointly assigning computation resources, uplink and downlink transfer power, and channel bandwidth, thus maximizing the overall energy gain (OEG) under the requirement of delay tolerance. Firstly, a computation offloading and energy harvesting model is built to capture CUs’ energy consumption, EUs’ energy harvesting, and time cost during offloading. Then, the problem is formulated as a mixed-integer, non-linear, and NP-hard programming problem. Considering its complexity, an iterative algorithm is proposed. It is proved that the result of the proposed algorithm approaches the optimal solution found by the method of exhaustion (MoE), while it is far less complicated. Simulations show that the proposed algorithm can significantly increase the OEG compared with the SNR-based offloading algorithm, random offloading, and no offloading. 

Further research on this topic is the introduction of service caching. The system model is modified to allow eRRHs to cache application services that are requested by CUs. If the required service is not cached in all eRRHs, this request will be forwarded to the cloud. This research intends to find and evaluate the influence of service caching on the OEG. After setting the caching strategy and different groups of requested services, a system model is built based on the previous model, and the proposed algorithm is also modified to match this new model. It is illustrated by the simulation that the proposed algorithm still works for service caching, and better performance is achieved when the similarity between the caching strategy and service request is high.